RESUMO
De novo variants in the Cytoplasmic FMR1-interacting protein 2 (CYFIP2) have been repeatedly associated with neurodevelopmental disorders and epilepsy, underscoring its critical role in brain development and function. While CYFIP2's role in regulating actin polymerization as part of the WAVE regulatory complex (WRC) is well-established, its additional molecular functions remain relatively unexplored. In this study, we performed unbiased quantitative proteomic analysis, revealing 278 differentially expressed proteins (DEPs) in the forebrain of Cyfip2 knock-out embryonic mice compared to wild-type mice. Unexpectedly, these DEPs, in conjunction with previously identified CYFIP2 brain interactors, included not only other WRC components but also numerous proteins associated with membraneless organelles (MLOs) involved in mRNA processing and translation within cells, including the nucleolus, stress granules, and processing bodies. Additionally, single-cell transcriptomic analysis of the Cyfip2 knock-out forebrain revealed gene expression changes linked to cellular stress responses and MLOs. We also observed morphological changes in MLOs in Cyfip2 knock-out brains and CYFIP2 knock-down cells under basal and stress conditions. Lastly, we demonstrated that CYFIP2 knock-down in cells, potentially through WRC-dependent actin regulation, suppressed the phosphorylation levels of the alpha subunit of eukaryotic translation initiation factor 2 (eIF2α), thereby enhancing protein synthesis. These results suggest a physical and functional connection between CYFIP2 and various MLO proteins and also extend CYFIP2's role within the WRC from actin regulation to influencing eIF2α phosphorylation and protein synthesis. With these dual functions, CYFIP2 may fine-tune the balance between MLO formation/dynamics and protein synthesis, a crucial aspect of proper mRNA processing and translation.
RESUMO
BACKGROUND: Molecular analysis of advanced tumors can increase tumor heterogeneity and selection bias. We developed a robust prognostic signature for gastric cancer by comparing RNA expression between very rare early gastric cancers invading only mucosal layer (mEGCs) with lymph node metastasis (Npos) and those without metastasis (Nneg). METHODS: Out of 1003 mEGCs, all Npos were matched to Nneg using propensity scores. Machine learning approach comparing Npos and Nneg was used to develop prognostic signature. The function and robustness of prognostic signature was validated using cell lines and external datasets. RESULTS: Extensive machine learning with cross-validation identified the prognostic classifier consisting of four overexpressed genes (HDAC5, NPM1, DTX3, and PPP3R1) and two downregulated genes (MED12 and TP53), and enabled us to develop the risk score predicting poor prognosis. Cell lines engineered to high-risk score showed increased invasion, migration, and resistance to 5-FU and Oxaliplatin but maintained sensitivity to an HDAC inhibitor. Mouse models after tail vein injection of cell lines with high-risk score revealed increased metastasis. In three external cohorts, our risk score was identified as the independent prognostic factor for overall and recurrence-free survival. CONCLUSION: The risk score from the 6-gene classifier can successfully predict the prognosis of gastric cancer.